Bruin
MCP server for AI agents to read, write, and organize notes in a local-first, human-in-the-loop note-taking app.
README
Bruin
The note-taking app built for AI agents — local-first, MCP-native, human-in-the-loop.
AI agents don't just answer questions. They should read, write, and organize your knowledge — with your approval.
Mac App Store · Direct Download · bruin.me
Direct download only: if macOS shows "app is damaged":
xattr -cr /Applications/Bruin.app
Quick Start — Claude Code / OpenClaw
Tell Claude this one sentence:
Read https://bruin.me/skills.md and help me install and set up Bruin.
Claude will read the skills file, walk you through adding the MCP config, and start using your knowledge base immediately.
Or set up manually in 2 steps:
1. Add to your MCP config (auto-writes the correct config):
/Applications/Bruin.app/Contents/MacOS/bruin --write-config
This writes to ~/.claude.json. The MCP server is built into the app — no npm, no Node.js required.
2. Install the Claude Code skill (makes Claude natively understand Bruin across all sessions):
/Applications/Bruin.app/Contents/MacOS/bruin --install-skill
That's it. Claude can now write notes, run daily logs, do research capture, build knowledge graphs, and more — all through natural conversation.
What Makes Bruin Different
| Traditional Note Apps | Bruin |
|---|---|
| AI is a sidebar chatbot | AI agents have 60 MCP tools — full CRUD, search, tagging, workflows |
| Notes are for humans only | Notes flow through draft → review → published — agents write, humans approve |
| Cloud-dependent | Local-first SQLite + iCloud sync — your data never leaves your machine |
| Flat file lists | Knowledge graph with [[wiki-links]], backlinks, force-directed visualization |
MCP Primitives
4 Prompts — Invoke directly in Claude Code
/mcp:bruin-notes:daily_log # Append to today's log
/mcp:bruin-notes:research_capture topic="AI frameworks" # Save research from any source
/mcp:bruin-notes:weekly_review # Summarize the week's notes
/mcp:bruin-notes:link_knowledge note_id="<id>" # Build wiki-links between notes
4 Resources — Reference with @bruin
| Resource | Contents |
|---|---|
bruin://notes |
All notes list |
bruin://notes/{id} |
Single note content |
bruin://tags |
All tags with counts |
bruin://daily |
Today's daily note |
60 Tools — Full reference at bruin.me/skills.md
| Category | Count | Key tools |
|---|---|---|
| Notes & Search | 16 | create_note, append_to_note, search_notes, semantic_search, advanced_query |
| Knowledge Graph | 3 | get_knowledge_graph, get_forward_links, get_backlinks |
| Agent Registry | 6 | register_agent, get_agent_audit_log, set_current_agent |
| Workspaces | 7 | create_workspace, bind_agent_workspace, switch_workspace |
| Tasks & Workflows | 10 | create_task, assign_task, create_workflow_template, execute_workflow |
| Webhooks | 6 | register_webhook, test_webhook, get_webhook_logs |
| Settings & Export | 5 | get_setting, export_note_markdown, export_note_html |
| Utility | 3 | pin_note, restore_note, list_tags |
Features
| Area | What you get |
|---|---|
| Editor | TipTap markdown with syntax highlighting, tables, images, slash commands |
| Search | FTS5 full-text + semantic search (all-MiniLM-L6-v2, auto-indexed) |
| Knowledge Graph | D3 force-directed visualization, backlinks, BFS traversal |
| Agent Registry | Register, track, and audit every agent that touches your notes |
| Task Management | Create, assign, and track tasks linked to notes and agents |
| Workflow Templates | Multi-step agent automation with variable chaining |
| Webhooks | HMAC-SHA256 signed callbacks with delivery logs and retry |
| iCloud Sync | Bidirectional sync with SHA-256 conflict resolution |
| 6 Themes | Dark Graphite, Charcoal, Solarized, Nord, Dracula, and more |
Architecture
AI Agent ──MCP (stdio)──► MCP Server (Node.js) ──► SQLite ◄── Rust Backend ◄──IPC──► React Frontend
│
iCloud Sync
(bidirectional .md files)
All three processes share the same SQLite database at ~/Library/Application Support/com.bruin.app/bruin.db. The MCP server notifies the Tauri app of changes via a trigger file that the file watcher detects.
| Layer | Technology |
|---|---|
| Desktop Shell | Tauri 2 |
| Backend | Rust, SQLite (WAL + FTS5), rusqlite |
| Frontend | React 19, TipTap, Zustand, D3 |
| Agent Protocol | MCP SDK (TypeScript) — Tools + Resources + Prompts |
| Embeddings | @huggingface/transformers (all-MiniLM-L6-v2, local) |
| Sync | iCloud Drive + notify file watcher |
Development
Prerequisites
- Node.js 18+
- Rust (stable) via rustup
- Xcode Command Line Tools
Setup
npm install
cd mcp-server && npm install && cd ..
npm run tauri dev # Vite + Tauri hot reload
Build
npm run tauri build # → .dmg in src-tauri/target/release/bundle/
Tests
npm test # Frontend (Vitest)
cd mcp-server && npm test # MCP server (96 tests)
cd src-tauri && cargo check # Rust type check
cd src-tauri && cargo clippy # Rust lint
Roadmap
- [ ] iOS / iPadOS companion app (read + quick capture)
- [ ] Collaborative workspaces (multi-user, shared knowledge graphs)
- [ ] Plugin system for custom MCP tool bundles
- [ ] PDF / web clipper import with auto-tagging
- [ ] End-to-end encryption for synced notes
Releasing
Version is tracked in three files (keep in sync): package.json, src-tauri/Cargo.toml, src-tauri/tauri.conf.json.
git tag v1.0.0
git push origin master --tags
GitHub Actions builds .dmg for Apple Silicon and Intel, then auto-publishes to Releases. See CHANGELOG.md for version history.
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。